On my projects at Target, I've had a good experience using DuplicateRecordFields coupled with a library like overloaded-records or generic-lens. This lets us define records that share field names without needing to use prefixes or multiple modules, while still being able to access and modify those fields easily thanks to the OverloadedLabels extension.

With this approach, you get a lens for every single record field without name conflicts. If you have a bunch of records with a field called id, you'll be able to access them using #id as a lens without running into name conflicts with other records *or* with normal identifiers (like the id function). You can then access fields with `record ^. #fieldName`, set them with the .~ operator and compose them with other lenses and traversals.

When we first out together a system like this I was worried that it would break type inference and cause weird error messages, but it had not caused *any* problems in practice. Haskell's records went from being a royal pain to being at least as convenient as any other language I've used, and sometimes more (eg when combined with traversals).

On Wed, Jan 29, 2020, 06:13 Jean Lopes <hawu.bnu@gmail.com> wrote:
Hi!

I'm trying to encode a domain which is full of simple records, and a lot of field name clashes.

Here is a simple example, a.k.a. 1:1 with the domain-spec:

data C001 {
  -- other stuff
  indMov :: Bool
}
data C170 {
  -- other stuff
 indMov :: Bool
}

There is ~148 records with this field (indMov), all with the same meaning. What are my options to encode this in Haskell?

- Prefix all record fields to avoid name clashes?
- Use DuplicateRecordFields?
- Typeclasses???
- other options I am not aware of?

which would you prefer? and why of course :)

Thanks,
Jean Lopes


 
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.